What do you think of the game AI?

DrClever

I'm not sure how often it has happened, if ever, that the game AI has been changed in any way but it has always been quite bad at playing the game and in some ways has become even worse as other aspects of the game have changed.

Are you happy with it the way it is?

entrailbucket

It's a good question, but ultimately the game doesn't actually work unless the AI is awful at playing it.

All the actual strategies in the game assume that you win 99.99% of the time on offense.

meadowsweet

Bless its stupid, familiar skull full of cotton socks.

I'm too lazy to dig it up, but years ago there was a great blog post by a developer at the time about the AI defense. The gist of what I remember is that they tried to improve it to play more optimally, and players really, really hated it because they were losing and using health packs way more often.

I mean, even today with all its faults, you have players swearing that it's cheating. We've all had those moments where it matched the tiles we needed to win even though their team doesn't even have powers in that color, or the conveniently random cascade that just happens to take out all those countdown tiles we just spread throughout the board.

It's a double-edged sword though. My Apocalypse is less valuable because I can't count on the AI to prioritize his powers properly on defense. But I also don't have to play against everyone else's optimized Apocalypses. Which in turn just elevates the characters with "no wrong answers" passive powers that the AI cannot screw up to be the new OP meta.

KGB

I love nothing more than pulling the wings off flies

@meadowsweet said:
I'm too lazy to dig it up, but years ago there was a great blog post by a developer at the time about the AI defense. The gist of what I remember is that they tried to improve it to play more optimally, and players really, really hated it because they were losing and using health packs way more often.

I remember that post.

It's why I've advocated for several years now that there should be 2 versions of the AI. The familiar one in PvE with the familiar rules.

The smarter one (as best they can make it) in PvP with the tradeoff that if you win, all characters are healed to 100% (no health pack issues) and if you lose (or quit) all 3 characters are downed. So in PvP all that matters is winning, no need to prioritize reducing damage or worrying if you lose a character or 2 during battle.

This makes PvP play entirely differently and would freshen up the game for those who have long been clamoring for something new game wise.

KGB

entrailbucket

@KGB said:

@meadowsweet said:
I'm too lazy to dig it up, but years ago there was a great blog post by a developer at the time about the AI defense. The gist of what I remember is that they tried to improve it to play more optimally, and players really, really hated it because they were losing and using health packs way more often.

I remember that post.

It's why I've advocated for several years now that there should be 2 versions of the AI. The familiar one in PvE with the familiar rules.

The smarter one (as best they can make it) in PvP with the tradeoff that if you win, all characters are healed to 100% (no health pack issues) and if you lose (or quit) all 3 characters are downed. So in PvP all that matters is winning, no need to prioritize reducing damage or worrying if you lose a character or 2 during battle.

This makes PvP play entirely differently and would freshen up the game for those who have long been clamoring for something new game wise.

KGB

I think it's probably too late for a change this big.

I'm not opposed to what you've described but it's a really, really big change from what we have, that'd have a bunch of knock-on effects (at the very very least you'd have to rework the entire reward structure, and I think some other mechanics would need changing too).

kuntilanak

It's not about how smart/dumb the AI is, but more on what the player loses after the match, whether winning or loosing. Considering this game is all about speed to either get the best possible rewards or minimize collateral, several 'meta' teams are formed to meet that criteria. Everyone knows that whenever cascade occurs, it takes forever to resolve, losing precious times.

I tend to plan my matches before hand, forming teams, selecting boosts and choosing opponents with highest chance of winning against. If a cascade crippled your character right before you win the match, you may need health pack to continue. What if you lose the match? You lose resources and time. But the AI loses nothing except some points in PvP. I believe that's the source of frustration. Worst is when you planned and boosted properly but loses to AI opening match/ cascade that left your guys severely crippled or AI gained tonnes of AP and wipe you in the next turn.

I agree with what @KGB said that they tweak the system so player emerges in full health if winning in PvP, but I'd say the opposite could work as well, meaning we lose but gain full health, and whatever damages received in matches stay if we win. Just ranting/spit-balling here...

entrailbucket

@KGB just out of curiosity, why would you want PvP to get the intelligent AI and PvE to stay dumb?

I'd push for the opposite -- PvE currently has zero external/meta strategy to keep it interesting, it's just "do super boring fights the fastest." PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

Phumade

TBH. I want the AI to randomly chose its color matches vs prioritizing its colors. That would intro a new element of board prediction vs knowing that the AI will take colors you need etc... People aren't asking for the game AI to be strategically playing for match4/5 2 drops into the future. They just do't want to see the AI prioritizing and stealing shared color ap. Maybe that's the real implementation. Have the AI randomly choose its color matches vs prioriting based on ability colors etc...

entrailbucket

@Phumade said:
TBH. I want the AI to randomly chose its color matches vs prioritizing its colors. That would intro a new element of board prediction vs knowing that the AI will take colors you need etc... People aren't asking for the game AI to be strategically playing for match4/5 2 drops into the future. They just do't want to see the AI prioritizing and stealing shared color ap. Maybe that's the real implementation. Have the AI randomly choose its color matches vs prioriting based on ability colors etc...

I've played tens of thousands of MPQ fights and have never noticed the AI prioritizing any color over another. He does love team up tiles for some reason, but aside from that I'm pretty sure that it's random.

Edit: I've heard this a lot over the years, though, so I asked in the dev questions thread! Let's hope they answer.

Phumade

@entrailbucket said:

@Phumade said:
TBH. I want the AI to randomly chose its color matches vs prioritizing its colors. That would intro a new element of board prediction vs knowing that the AI will take colors you need etc... People aren't asking for the game AI to be strategically playing for match4/5 2 drops into the future. They just do't want to see the AI prioritizing and stealing shared color ap. Maybe that's the real implementation. Have the AI randomly choose its color matches vs prioriting based on ability colors etc...

I've played tens of thousands of MPQ fights and have never noticed the AI prioritizing any color over another. He does love team up tiles for some reason, but aside from that I'm pretty sure that it's random.

It pretty sure it does. If the AI teams primary colors are Black Yellow Red for AP gen. It will prioritize those colors over blue green pink. and vice versa if its primary colors are different. but yeah, it will prioritize a match 4 even in a weak color over a primary color that enables a power.

For sure though, it depends on how you define prioritze.

Having it explicitly seek the non primary colors would be a good way to dumb down the AI

skittledaddy

meadowsweet said:
I'm too lazy to dig it up, but years ago there was a great blog post by a developer at the time about the AI defense. The gist of what I remember is that they tried to improve it to play more optimally, and players really, really hated it because they were losing and using health packs way more often.

I mean, even today with all its faults, you have players swearing that it's cheating. We've all had those moments where it matched the tiles we needed to win even though their team doesn't even have powers in that color, or the conveniently random cascade that just happens to take out all those countdown tiles we just spread throughout the board.

It's a double-edged sword though. My Apocalypse is less valuable because I can't count on the AI to prioritize his powers properly on defense. But I also don't have to play against everyone else's optimized Apocalypses. Which in turn just elevates the characters with "no wrong answers" passive powers that the AI cannot screw up to be the new OP meta.

I think this is the link to the post, but it appears to no longer be available.
[https://www.demiurgestudios.com/engineering/a-worthy-opponent/]

KGB

I love nothing more than pulling the wings off flies

@Phumade said:

@entrailbucket said:

@Phumade said:
TBH. I want the AI to randomly chose its color matches vs prioritizing its colors. That would intro a new element of board prediction vs knowing that the AI will take colors you need etc... People aren't asking for the game AI to be strategically playing for match4/5 2 drops into the future. They just do't want to see the AI prioritizing and stealing shared color ap. Maybe that's the real implementation. Have the AI randomly choose its color matches vs prioriting based on ability colors etc...

I've played tens of thousands of MPQ fights and have never noticed the AI prioritizing any color over another. He does love team up tiles for some reason, but aside from that I'm pretty sure that it's random.

It pretty sure it does. If the AI teams primary colors are Black Yellow Red for AP gen. It will prioritize those colors over blue green pink. and vice versa if its primary colors are different. but yeah, it will prioritize a match 4 even in a weak color over a primary color that enables a power.

For sure though, it depends on how you define prioritze.

Having it explicitly seek the non primary colors would be a good way to dumb down the AI

Agreed. The AI prioritization seems to be:
1) Match 5
2) Match 4
3) Match in a primary color or TU (until the TU is fired after that it stops caring about TU)
4) Non-primary color.

I think the reason it appears random to EB is that because he fights a lot of rainbow teams that have all 6 colors as primary (spread across 3 enemies). But take a closer look when you fight a team that isn't rainbow or when 1+ enemies is downed and there is now a color that is no longer primary and you'll notice the AI ignores it.

The other place to easily see this is Boss (or 1v1) fights. Bosses have at most 3 primary colors and ignore the other 3.

KGB

KGB

I love nothing more than pulling the wings off flies

@entrailbucket said:
@KGB just out of curiosity, why would you want PvP to get the intelligent AI and PvE to stay dumb?

I'd push for the opposite -- PvE currently has zero external/meta strategy to keep it interesting, it's just "do super boring fights the fastest." PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

I personally (and I presume most people like yourself) consider PvE to be like riding a bike with training wheels or a beginner mode. A lot of that is because there are a ton of non-tile movers that you fight (some PvE's are almost entirely non-tile movers). Plus I presume most new players play PvE long before they venture into PvP (it's a natural place to move to after doing the prologue). So it makes sense to me to keep the AI weaker in that mode so as not to scare away new players.

KGB

entrailbucket

@KGB said:

@entrailbucket said:
@KGB just out of curiosity, why would you want PvP to get the intelligent AI and PvE to stay dumb?

I'd push for the opposite -- PvE currently has zero external/meta strategy to keep it interesting, it's just "do super boring fights the fastest." PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

I personally (and I presume most people like yourself) consider PvE to be like riding a bike with training wheels or a beginner mode. A lot of that is because there are a ton of non-tile movers that you fight (some PvE's are almost entirely non-tile movers). Plus I presume most new players play PvE long before they venture into PvP (it's a natural place to move to after doing the prologue). So it makes sense to me to keep the AI weaker in that mode so as not to scare away new players.

KGB

That's what the prologue is for!

What if the AI changed based on CL instead?

Omegased

your responses are too confusing.

I like the AI as it is. making it smarter will mean making it harder and sad.

DrClever

@entrailbucket said:
PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

Most of this metagame is for me the absolute worst thing about MPQ and if I had my way all development effort would be directed towards eliminating it and its PvE equivalents.

I hate the idea that playing the actual game is the last thing players want to be doing and that winning should be trivial and quick.

Give me a strong AI and sense of achievement from getting a win, please. (I realise this isn't the way to generate revenue and so won't happen.)

I guess what I'm saying is that I should find something else to play.

Omegased

@DrClever said:

@entrailbucket said:
PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

Most of this metagame is for me the absolute worst thing about MPQ and if I had my way all development effort would be directed towards eliminating it and its PvE equivalents.

I hate the idea that playing the actual game is the last thing players want to be doing and that winning should be trivial and quick.

Give me a strong AI and sense of achievement from getting a win, please. (I realise this isn't the way to generate revenue and so won't happen.)

I guess what I'm saying is that I should find something else to play.

To be fair tho - considering the passive nature of meta characters, there's barely any "skill" anyway

bluewolf

Bless its stupid, familiar skull full of cotton socks.

@DrClever said:

@entrailbucket said:
PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

Most of this metagame is for me the absolute worst thing about MPQ and if I had my way all development effort would be directed towards eliminating it and its PvE equivalents.

I hate the idea that playing the actual game is the last thing players want to be doing and that winning should be trivial and quick.

Give me a strong AI and sense of achievement from getting a win, please. (I realise this isn't the way to generate revenue and so won't happen.)

I guess what I'm saying is that I should find something else to play.

If you do, I for one would miss your forum posts, which often make me chuckle.

entrailbucket

@DrClever said:

@entrailbucket said:
PvP at least has coordination, sniping/dumping, hopping, switching teams, etc to keep it interesting.

Most of this metagame is for me the absolute worst thing about MPQ and if I had my way all development effort would be directed towards eliminating it and its PvE equivalents.

I hate the idea that playing the actual game is the last thing players want to be doing and that winning should be trivial and quick.

Give me a strong AI and sense of achievement from getting a win, please. (I realise this isn't the way to generate revenue and so won't happen.)

I guess what I'm saying is that I should find something else to play.

Yeah...like, I don't disagree with you. I just don't know how, at this point in the game's lifecycle, they can actually do anything meaningful about any of that stuff without essentially making a new game. It's not about revenue as much as it is a total rethink/rebuild of every single mechanic.

meadowsweet

Bless its stupid, familiar skull full of cotton socks.

@skittledaddy said:
I think this is the link to the post, but it appears to no longer be available.
[https://www.demiurgestudios.com/engineering/a-worthy-opponent/]

Thanks for the assist @skittledaddy ! The Internet Archive Wayback Machine has us both covered:
https://web.archive.org/web/20170507034150/http://www.demiurgestudios.com/engineering/a-worthy-opponent

@KGB said:
Agreed. The AI prioritization seems to be:
1) Match 5
2) Match 4
3) Match in a primary color or TU (until the TU is fired after that it stops caring about TU)
4) Non-primary color.

It's actually even dumber than that... [from the blog link above]

The section that tallies up the number of tiles was written to check one tile behind the swap and two tiles ahead, so it doesn't look far enough for match‑5s in a line. Since it only checks along one axis, it can't see L and T shaped match‑5s.

This scoring system means that the criteria for selecting a swap are:

1. Number of matched tiles in a straight line, up to 4
2. Does the match charge an ability on the computer's team?

That's it. That's all the logic. The first rule takes precedence over the second: the computer will always take a match‑4 over a match‑3. If there are two match‑4s available, the computer will prefer the one that will charge its abilities. If there are multiple best swaps, it will randomly choose one. Sometimes the match-4 also results in a match‑5, but the computer doesn't “see” the match-5.

entrailbucket

Huh. Assuming that's still true, he doesn't actually prioritize his strongest colors directly -- he only prioritizes matches that will charge an active ability. So it's none of the things any of us thought!